Processing symplectic methods for near-integrable Hamiltonian systems (Q1841530)

The authors study processing symplectic methods for Hamiltonian systems, i.e. the methods of the form \(e^{-hH}=e^Pe^{-hK}e^{-P}\), where \(P\) is called the processor and \(K\) the kernel. The basic idea behind such methods is that its \(N\)-th power is \(e^P(e^{-hK})^Ne^{-P}\), so that the processor must be evaluated only when output is desired. The Hamiltonian of a near-integrable system splits naturally into two parts \(H=H_0+\varepsilon H_1\) with a perturbation parameter \(\varepsilon\). The kernel \(K\) and the processor \(P\) may now be constructed by compositions of the flows of \(H_0\) and \(H_1\) via the Baker-Campbell-Hausdorff formula in order to get symplectic methods. The authors study the arising order conditions and determine optimal solutions for them. The obtained methods are applied to some concrete examples and compared with symplectic methods without processing. The results demonstrate the efficiency of the processing approach.